Download Report

REVIEW OF SCIENTIFIC INSTRUMENTS 78, 095107 共2007兲
Enlarged atomic force microscopy scanning scope: Novel sample-holder
device with millimeter range
A. Sinno, P. Ruaux, L. Chassagne, S. Topçu, and Y. Alayli
LISV, University of Versailles, 45 Avenue des Etats Unis, 78035 Versailles, France
G. Lerondel, S. Blaize, A. Bruyant, and P. Royer
Laboratoire d’Instrumentation Optique et de Nanotechnologies, ICD, CNRS FRE 2848, Université
de Technologie de Troyes, 12 Rue Marie Curie BP2060, 10010 Troyes, France
共Received 15 May 2007; accepted 29 July 2007; published online 27 September 2007兲
We propose a homemade sample-holder unit used for nanopositionning in two dimensions with a
millimeter traveling range. For each displacement axis, the system includes a long range traveling
stage and a piezoelectric actuator for accurate positioning. Specific electronics is integrated
according to metrological considerations, enhancing the repeatability performances. The aim of this
work is to demonstrate that near-field microscopy at the scale of a chip is possible. For this we chose
to characterize highly integrated optical structures. For this purpose, the sample holder was
integrated into an atomic force microscope. A millimeter scale topographical image demonstrates
the overall performances of the combined system. © 2007 American Institute of Physics.
关DOI: 10.1063/1.2773623兴
I. INTRODUCTION
With the development of nanotechnologies, scanning
over large areas with a nanometer resolution is becoming
more and more important. However, millimeter range scanning is not yet commercially available. This is because nanometer specifications such as resolution, positioning accuracy, and repeatability of the displacements are to meet all
along the surface to cover. At the present time, with commercial systems, displacement is either limited to about a hundred of micrometers or possible on the millimeter scale but
with limited accuracy and repeatability 共typically in the order
of the micrometers兲. Furthermore, repeatability is often neglected restricting the scope of applications. The technological progress of industrial apparatus makes us think this will
be a serious challenge in the near future.1 Yet, cutting-edge
technologies are seeking new frontiers: microelectronics,
near-field characterization of large scale structures, and electronic lithography for very large scale patterning, for example. The concepts of traceability, accuracy, and repeatability are thus essential. In this work, we report on a
displacement system with a millimeter range and a nanometric positioning repeatability and traceability in both X and Y
directions. In addition we show that this system is compatible with a commercially available atomic force microscopy
共AFM兲 head.
This study was initially targeted toward applications in
scanning probe microscopy 共SPM兲. Surface inspection of optical components will require millimeter long scans along the
light propagation axis. This is the case of lightwave device
characterization where a combined AFM for topography imaging and a scanning near-field optical microscope 共SNOM兲
for optical inspection is used. The aim is to observe the emitting or internal field of a structure according to its geometry.2
Such a scan may reveal relevant information such as the
0034-6748/2007/78共9兲/095107/7/$23.00
phase and amplitude of the profiles of the guided field and
from there, the mode effective indexes or refractive index
profiles and local diffraction losses. When combined with
Fourier transform, long scans become necessary to increase
the precision of the refractive index measurement.3 At the
time being, SPM scans are only possible over 100 ␮m,
which is the piezoelectric actuator 共PZT兲 limit. Moreover,
the electronic bandwidth of the PZT is quite low in these
apparatus. The resolution is also limited by the acquisition
system, which limits the number of points to typically 1024
in both directions. In this case, scanning over the full range
of 100 ␮m leads to a poor resolution of about 100 nm.
Nanometric resolution can thus be obtained only for micrometric scans.
The method currently used to overcome the 100 ␮m
limit consists in stitching together overlapping elementary
microscopic 2D scans.3 A typical overlapping percentage of
about 30% on each side of the scanned area is necessary. The
overlapping scans are afterwards glued together by means of
image processing techniques based on the localization of distinguishable topography patterns or defects that would match
on the overlapping area. This process is time consuming and
limited to the presence of a significant number of signatures
on the surface that would allow a precise matching. For these
reasons, this method is rarely used. Another solution is to use
an array of probes. In Ref. 4, an image of 2 ⫻ 2 mm2 was
obtained but with a pixel size of 400 nm. This method requires to synchronously acquire the data from each probe, in
addition to precisely mount the array of probes.
For positioning over millimeter ranges with high resolutions, the solution which appeared recently is the coupling of
a mechanical displacement stage for long range motion with
a precision stage for accurate positioning.5–7 Still under development for years to come, several metrology institutes in
the world have launched projects of sample-holder systems
78, 095107-1
© 2007 American Institute of Physics
Downloaded 03 Jun 2008 to 193.51.41.209. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp
095107-2
Sinno et al.
FIG. 1. 共Color online兲 Sketch of the system. The double mechanical stage is
composed of the XY linear motors and an XY piezoelectric actuator. Two
interferometers allow real time position monitoring. The Personal computer
共PC兲 unit controls the linear motors and the homemade electronic board that
generates the phase-shift commands to control the piezoelectric actuators.
AFM acquisition is synchronized with the XY displacements.
with nanometric resolution and accuracy over millimeter
ranges: NIST 共National Institute of Standards and Technology兲 in the USA,8,9 PTB 共Physikalisch-Technische Bundesanstalt兲 in Germany,10 Metas in Switzerland,11 NPL 共National Physic Laboratory兲 in the United Kingdom,12,13 NIM
共National Institute of Metrology兲 in China, and LNE 共Laboratoire National de métrologie et d’Essais兲 in France.14 Furthermore, many works were initiated to develop an apparatus
usually named metrological atomic force microscope15–21
共MAFM兲, but most of them still present some limitations
with the traceability and the reproducibility of the imaging
process; hysteresis and nonlinearity defects of the piezoelectric actuator induce problems of repeatability of the motion.
It seems very difficult to cross the 100 nm barrier in accuracy without compensation.22 First, apparatus integrating
these capabilities are being developed in laboratories. A first
demonstration of a long scan was illustrated in Ref. 10, but
with a resolution of only 100 nm and a very long time consumption budget where 12 h were necessary for a scan of
1 mm⫻ 100 ␮m.
This article presents the system we proposed for precision scanning over millimeter range with high resolution. It
is divided into two main sections. The first one is devoted to
the scanning stage where after presenting the stage, hysteris,
mechanical, software and scanning methods issues are addressed. In the second section, imaging results are presented.
The overall performances of the combined system are finally
discussed.
Rev. Sci. Instrum. 78, 095107 共2007兲
displacement axis. The first stage is a commercially available
long range linear motor 共Aerotech ANT-50兲 capable to move
over 50 mm with a 10 nm resolution and a straightness of
about 1 ␮m. The second stage comprising a piezoelectric
actuator 共PiezoSystem Jena兲 overcomes the defects of the
first stage thanks to its subnanometric resolution and high
bandwidth. Its full range is 15 ␮m but a low voltage amplifier limits its range to 3 ␮m. Position measurements are
made with a frequency stabilized HeNe laser that ensures
traceability and a double two-pass Michelson interferometer
with a resolution of 0.31 nm. Such interferometers have a
nonlinearity of 2 nm if finely adjusted.24 A homemade high
frequency electronic board equipped with a phase-shifting
electronic 共PSE兲 circuit acts as the core of the system.25 It
generates high frequency signals used as reference signals to
synchronize the laser head and its measurement boards, the
Zygo ZMI2001 boards. The PSE circuit controls the PZT
actuators with subnanometric position steps by imposing
well calibrated phase shifts between the reference signal and
the measured signals. The performances of the electronic
board in terms of positioning resolution and repeatability are
at the nanometer scale. The phase shifts ⌬␾ are configurable
so that downscaled position steps ⌬d can be achieved according to the equation
⌬d =
␭ 0⌬ ␾
,
8␲n
共1兲
where ␭0 is the wavelength of the laser source in vacuum and
n the refractive index of air. The laser source was calibrated
at the LNE. Its wavelength is equal to ␭0 = 632.991 528 nm
with a relative uncertainty of 1.6⫻ 10−9 共1␴兲. The displacement resolution can be programed down to 0.25 nm. This
phase-shifting principle frees our displacements from the
nonlinearity and hysteresis effects of the PZT actuator. Detailed description of the HF board can be found in Ref. 25. A
Zerodur® mirror cube acting as a moving mirror for the laser
interferometers is placed on top of the XY PZT actuator and
serves us to carry a sample on its top face, making of our
system a sample-holder unit 共SHU兲. The SHU is adapted
under an atomic force microscope 共M5 Park Scientific Instruments兲 allowing sample surface inspection beyond the
microscope’s area coverage limit. The SHU ensures the XY
displacements, whereas the AFM controller handles the
z-axis movements and the atomic force feedback. The z actuator of the microscope is a 7 ␮m range PZT scanner. The
tip of the microscope is brought in close proximity of the
sample surface. The tip-to-surface distance is controlled by
the atomic force regulation. The whole system is supervised
and controlled through a LABVIEW interface. The sketch of
the system is presented in Fig. 2 on the upper part and a
photography on the lower part.
II. EXPERIMENTAL SETUP
A. Sketch of the system
In previous works, we presented an optoelectronic system able to perform displacement steps over millimeter
ranges with nanometric accuracy and repeatability.23 In this
article, we extend the principle to a double axes system 共Fig.
1兲. This system includes a double moving stage for each
B. Hysteresis control
In classical AFM controllers, intrinsic nonlinearity defects of the PZT-driven scanner are software controlled. The
inverse hysteresis model is usually used10,26 according to the
equation
Downloaded 03 Jun 2008 to 193.51.41.209. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp
095107-3
Enlarged AFM scanning scope
Rev. Sci. Instrum. 78, 095107 共2007兲
effect on the bandwidth of the control loop. In our system,
nonlinearity rectification is hardware implemented based on
the phase locking principle and depends on the robustness of
the analog servo loop controller that is implemented.
C. Mechanical adjustments
FIG. 2. 共Color online兲 Top part: sketch and photography of the system. One
can see the mirror cube where the sample is to be placed, the 2D piezoelectric actuator, the mechanical stages used for fine angles adjustments, the
linear motors in addition to the AFM head on top of the displacement system. Tuning axes and angles are pointed out on the CAD design. 共Xo , Y o兲
optical axes, 共X P , Y P兲 piezoelectric actuator axes, and 共Xm , Y m兲 linear motors
axes must all be aligned previous to operation. Bottom part: photography of
the AFM tip placed over the SHU.
V共x兲 = x−1共V0,x0兲
共2兲
in order to calculate the desired control voltage V to apply to
the PZT actuators to attain a position x as a function of the
actual state 共voltage V0 and position x0兲 of the actuator. This
method presents two major drawbacks. The first one is that
this equation encloses linearization coefficients that were defined under specific environmental conditions and implemented on board of the AFM controller. These coefficients
and model will no more be valid if the conditions change and
the PZT behavior will consequently not be the same. PZT
actuators undergo aging with time and have a different behavior at each stable operating temperature. This makes the
AFM displacements inaccurate, unless real time calibration
is used. The SHU follows a different directive. The PZT in
our system tries to keep track of the desired trajectory by
locking the phase of the measured signal on the phase of the
reference signal. This means that the PZT will contract/
retract by the necessary distance until phase difference nulls,
whatever the actual state of the PZT is and whatever the
operating temperature at which the system has stabilized.
Hysteresis, creep, and temporal instabilities of the PZT are
eliminated which enhances the repeatability of the displacements. The second drawback is that Eq. 共2兲 is usually software implemented in the AFM controller. Such equation usually incorporates a lot of coefficients and thus needs complex
computation, especially when complex algorithms are used
to better track the linear model.27 This may have an annoying
Since the SHU operates over millimeter ranges, tiny mechanical defects are magnified and may become uncontrollable. Preliminary alignments are thus necessary to minimize
errors before operating the SHU. The orthogonality between
two perpendicular faces of the mirror cube is certified to be
less than 1 arc sec. The orthogonality between the two linear
motors was controlled using interferometry with an accuracy
of 5 arc sec. Furthermore, several rotational stages inserted
between the linear motors and the piezoelectric actuators allow angles and axes tuning 共Fig. 2兲. The optical axes of the
interferometers 共Xo , Y o兲, the mechanical axes of the linear
motors 共Xm , Y m兲, and the mechanical axes of the piezoelectric actuators 共X p , Y p兲 are finely aligned with an uncertainty
lower than 5 arc sec.28 The remaining straightness defects
can be compensated by the piezoelectric action using the
electronic feedback lock-in system. The sample to investigate is centered on the top face of the mirror cube and thus
practically on the intersection point of Xo and Y o for an Abbe
error-free configuration. Another error to overcome when
dealing with millimeter ranges is the tilt that may arise between the XY plane of the moving stages and that of the
sample being scanned. The greater the tilt and the longer the
distance covered, the more the PZT actuator of the microscope will have to contract/retract in the z direction in order
to track the surface of the sample. A microscope equipped
with a 10 ␮m range PZT actuator in the z axis will unlock
after traveling 1 mm in the XY plane if a tilt of 10 mrad
exists. This tilt causes an artifact giving the impression that
the two sample surfaces are not parallel to each other. As far
as mechanical vibrations are concerned, the interferometers,
mechanical stages, and AFM apparatus lay on an active antivibration table that eliminates external noise. Temperature
of the room is controlled in order to prevent any drift on the
laser measurements29 and mechanical parts.
D. Software adaptation and acquisition
AFM controllers present many limitations. With classical AFM software, the scans are programable with two major
dimensional parameters: the size of the scan 共in micrometers兲
and the number of points to sample for each line. This directly induces the spatial resolution of the image. The maximal number of points per line is usually limited to 1024. For
example, a scan of 100⫻ 100 ␮m2 displayed using 1024
⫻ 1024 pixels leads to a resolution of about 100 nm in each
direction. A scan of 1 mm long with 1024 sampled points
would lead to a resolution of 1 ␮m / pixel. For this reason,
we shunt the acquisition module of the AFM and take hand
over it in order to increase the resolution especially for long
range scans. A motion synchronized pulse firing controller
共MSPFC兲 triggers the acquisition from the AFM synchronously with the SHU movements. The MSPFC pulses are
derived from the signals of the optical encoders of the linear
Downloaded 03 Jun 2008 to 193.51.41.209. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp
095107-4
Sinno et al.
motors and from the homemade phase-shifting board. Each
time the SHU steps in the XY plane, the MSPFC launches an
interrupt pulse toward a LABVIEW real time based data acquisition card that will acquire the z-position signal from the
AFM. The analog to digital converter 共ADC兲 of the acquisition board is a 16 bit converter and operates at sampling
frequencies over 100 kHz. The MSPFC is able to output
pulses which allow us to ally high speeds with high resolutions without altering the accuracy. The number of data
points per line acquired is then only limited by the time
resolution of the pulse incoming from the MSPFC synchronization board. The real limitation is, in fact, a software
problem: the file system of the operating shell does not allow
huge file indexation. For this, the acquisition software stores
each set of lines in a file. A powerful computer able to handle
the high speed flow of data when the SHU runs at high
speeds and to process the resulting amount of data points is
also needed. Furthermore, classical AFM softwares display
the scanned surface in real time. This dramatically decreases
the bandwidth of the acquisition since image nonlinearity
corrections, among others, are processed online. Furthermore, the link between the AFM controller and the host computer used for display is usually an Ethernet connection30
with a limited bandwidth that is not suitable for real time
display of high speed data. In our system, data processing
and displaying are made afterwards in order to visualize the
three-dimensional 共3D兲 topography image. The method we
use produces data files containing the entire information, and
any image processing script can be applied to reveal information about the sample such as artifact elimination or outlining specific features of the topography.
E. Scanning methods
Traditional AFM tips scan the surface of a sample in a
zigzag trajectory mode by applying sweeping command signals on the piezoelectric actuators. A sweep is made to scan
a line along the X axis, and a second slower sweep is made
along the Y axis to compose the multiline profile. The SHU
operates in a different manner. The XY controllers of the
AFM are unused since the SHU is ensuring the XY displacements. The tip of the microscope is equipped with a PZT
actuator that is acting only in the z direction. We then figured
out three functional modes with different scanning strategies
in order to move the sample relative to the tip over the millimeter range.
In the first mode, named “matrix by matrix” 共MM
mode兲, a multitude of elementary micrometric scans called
matrices is achieved and jointed to compose the whole
image—the mother matrix. An initialization phase is first
made to mark a 共0; 0兲 position in the 共X ; Y兲 reference. Then
a first elementary matrix is scanned by the piezoelectric actuators on micrometric range with preprogramed parameters
such as speed, elementary area, and resolution. Elementary
matrix area is limited to 3 ⫻ 3 ␮m2 since achieved using the
PZT actuators only. Once this first scan is achieved, the PZT
actuators which attained their stretching limit are released.
The linear motors then move on to displace the sample and
position it in a way a new elementary matrix can be scanned
accurately next to the previous one. The piezoelectric actua-
Rev. Sci. Instrum. 78, 095107 共2007兲
tors can hence achieve several scans that are perfectly adjacent until an image composed of several elementary matrices
is formed. Under MSPFC control, the acquisition points are
synchronized with the phase-shifting system which ensures
to be insensitive to the nonlinearities and defects of the piezoelectric actuators. The MSPFC launches an interrupt toward the AFM each time the PZT steps inside an elementary
matrix. This mode resembles to the stitching process but no
stitching is applied a posteriori. The phase-shifting positioning system ensures the accuracy of the connections between
the elementary matrices; overlapping windows are not
needed. The position uncertainty is programable, depending
on the resolution of the positioning system which can be as
low as 0.25 nm.25 This mode is sufficiently accurate but its
problem is that it is time consuming.
In the second mode, named “line per line” 共LL mode兲,
long lines are performed with the linear motors without piezoelectric compensation along the displacement axis. Only
the straightness of the moving motor is controlled. The same
initialization phase is made fixing a 共0;0兲 position in the
共X ; Y兲 reference. Then, a long line is directly performed by
the linear motor along the X axis. The piezoelectric stage
only compensates the straightness along the Y axis, but not
the accuracy position defects along the X axis. When a line is
done, the phase-shifting system repositions the sample for a
second line strictly parallel to the previous one. This routine
is applied until covering the whole area to scan. The resulting image is then accurate in the Y dimension. The length of
the lines and the spacing between two lines are programable.
The displacement resolution along the X axis is 10 nm since
the linear motor is traveling alone without the help of the
PZT stage. The MSPFC synchronizes the AFM acquisitions
with the displacements of the linear motor. This mode is not
optimized for position control but allows long scans easily
and quickly. Speeds at the mm/s scale can be reached.
In the third mode, called “tracking trajectory mode” 共TT
mode兲, long lines are performed using the linear motors with
real time piezoelectric compensation permanently activated
on both axes. The linear motors and the PZT actuators cooperate together to travel long lines accurately. Each time the
linear motor moves a step, the PZT actuator compensates its
defects ensuring the positioning accuracy. This implies lower
speeds. In fact, the PZT actuators have an open loop bandwidth of 2.5 kHz, but the closed loop control and the voltage
amplifier scale it down to less than 1 kHz. Tests showed that
a speed of 10 ␮m / s is the maximum value so that the PZT
actuators can track. At the end of each line, the PZT actuators will have contracted/retracted by the amount of nonlinearities, straightness, hysteresis, and other intrinsic defects of
the system that were cumulated all along a line. The same
outcomes are then inherited from the LL mode: straight and
accurately distanced long lines, but in addition to this, intrinsic defects of the linear motors are compensated at each
point. The TT mode is the ultimate mode since accurate in
both dimensions but lacks of speed in addition to being the
most difficult to optimize for all the lock-in loops involved in
this type of mode control.
Downloaded 03 Jun 2008 to 193.51.41.209. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp
095107-5
Enlarged AFM scanning scope
Rev. Sci. Instrum. 78, 095107 共2007兲
III. MEASUREMENTS AND DISCUSSIONS
A. Parameters of the experiment
The sample investigated is a silicon-on-insulator 共SOI兲
substrate with a waveguide structures pattern. The
waveguides are parallel but irregularly set. Their width is
about 2.8 ␮m and their height around 160 nm. In order to
minimize the total scanning time, the speed of the scan has to
be maximized. For this purpose, the LL mode is the most
convenient. A speed of 40 ␮m s−1 was chosen. The SHU
allows higher speeds but this upper limit is fixed by the AFM
operation regulations: damage may, in fact, be caused to the
AFM tip if this speed is larger. The AFM is functioning in
tapping mode. This implies that the signal that represents the
surface profile is the voltage applied to the z-PZT actuator to
keep constant the distance from the sample and null the error
signal of the AFM control loop. It is this signal that is sent to
the LABVIEW acquisition card for sampling, triggered externally with the pulses coming from the MSPFC that are synchronous with the long axis displacements. For a scan of
1 mm long, each line takes 25 s, but the positioning process
at nanometric accuracy when the SHU is changing the line
can be time consuming and takes around 30 s, so the total
time for a line scan is around 1 min.
The image presented in Fig. 3 is 750 nm width 共X axis兲
and 1 mm long 共Y axis兲. The resolutions are 25 nm for the X
axis 共29 lines兲 and 20 nm for the Y axis. So there is 50 000
points per line along the Y axis. This is not limited by the
apparatus or by the acquisition system, but we preferred to
limit the number of points to make the data treatment easier.
The Z axis represents the AFM topography signal.
B. Results for long topography imaging
Figure 3 was obtained using MATLAB software. The upper part of the figure is presented without any post-treatment
共raw image兲. Note that the long scan axis is Y, whereas X
stands for the short scan direction. We can notice shading off
of colors in the image because of the remaining tilts in X and
Y axes. Due to the data high density, the image is not really
easy to read. One can see the parallel waveguide structures
along the X axis and their irregular layout. The lower part
shows the same image but after background correction. Here,
the background is mainly due to a residual tilt between the
sample and the top face of the cube and a thermal drift during the acquisition process. A thermal drift acts to expand the
metrological chain of the system. This chain includes the
body of the microscope in addition to the multistacked elements of the SHU.
In order to analyze the data along the long scan axis Y, a
cross section is pointed out in Fig. 4. The upper part is presented, also without any post-treatment 共raw profile兲. The
waveguide structures clearly appear here. They seem to be
distorted, but it is only an artifact due to the disproportional
representation of high density data on the Y axis with respect
to the X axis. This cross section was flattened using a third
order polynomial fit and is shown in the lower part of the
figure. The equation used for the background correction is
Z = −390+ 1.96Y − 7.67⫻ 10−7Y 2 + 3.3⫻ 10−13Y 3, where Y is
expressed in micrometers. This correction may have several
FIG. 3. 共Color online兲 Long topographic image. This image is 750 nm wide
共X axis兲 and 1 mm long 共Y axis兲. The resolutions are 25 nm for the X axis
共29 lines兲 and 20 nm for Y axis 共50 000 points兲. The Z axis is the acquisition
of the AFM signal. The entire image was achieved in about 25 min at a scan
speed of 40 ␮m s−1. The upper part is presented without any treatment. We
can see shading-off colors due to remaining tilts in X and Y axes; the
waveguides are rather difficult to detect. The lower part is the same image
after background adjustment; waveguide structures appear along X axis. The
image reveals that on this sample, the structures are parallel but irregularly
spaced 共according to the design pattern兲.
origins. These include a vertical tilt between the translation
stage and the sample, an eventual thermal drift along the z
axis,10 in addition to a slight curved discrepancy due to either
flatness defects of the SHU and the radius of curvature of the
SOI sample. The waveguide structures seem distorted but
this is still an artifact. One can note that the pedestals are
asymmetrical. This artifact is due to scanning at a very high
lateral speed of 40 ␮m / s. It can be eliminated by the use of
a convenient tip in addition to subtracting the effect of abrupt
tip-sample interactions.
Figure 5 shows a zoom revealing a single structure at the
Y position of 217 ␮m. The upper part presents a 3D view in
order to show a waveguide profile, while the lower part presents a cross section which allows appropriate measurements. With this zoom, the waveguide structure is not distorted anymore. The height and the width are compatible
with those expected.
Downloaded 03 Jun 2008 to 193.51.41.209. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp
095107-6
Sinno et al.
Rev. Sci. Instrum. 78, 095107 共2007兲
sentation of the parameters of both the macro- and nanostructures. With the millimeter large image of Fig. 3, a lot of
waveguides are included. It is hence possible to check out
the homogeneity at the chip level by means of statistical
measurements.
C. Discussions
FIG. 4. Cross section of Fig. 3 along the Y axis. The upper part is presented
without any treatment and points out the residual tilt. On the lower part, a
third order polynomial fit was calculated and subtracted. The tilt that was
subtracted corresponds to several cumulated effects: mainly the residual tilt
but also the radius of curvature of the silicon substrate, slight flatness defects
of the translation stages, and a probable thermal drift. Discrepancies at the
base of the waveguide structure are artifacts due to the tip shape.
It is important to note that this zoom was obtained directly from the original image, while with a conventional
AFM system, it would have been necessary to scan once
again. The resolution is good enough to have a good repre-
The herein reported millimeter scale atomic force microscopy image was obtained by 1 mm single scans repeated
29 times with a 25 nm spacing. Although accurate in only
one dimension, the line-per-line mode was used. This first
result demonstrates the feasibility of large displacements
while addressing the constraints related to two-dimensional
共2D兲 large scale imaging. The LL mode allows, in fact, us to
scope large areas easily and quickly. However, results
showed the expected layout of the structure that was
scanned. No XY drift was noticed all along the surface. Accuracy in the z direction is left to the AFM and depends on
the sensitivity of the sensors the AFM uses. In this experiment, a quadquadrant photodetector associated with an optical lever arm scheme was used. To meet higher performances, the SHU will have to be endowed with a traceable
measurement along the z axis. This can be obtained by integrating a third laser interferometer within the SHU system.
Moreover, environmental condition supervision is critical to prevent any drift on the resulting image. For the SHU
measurements to be more accurate, drift supervision and
control shall be incorporated.10 Refractive index measurements of the ambient air can also be taken into account with
Édlen formula31 to improve the positioning accuracy.
In addition to this, we plan in future works to extend the
imaging area covered by the SHU especially in the short
scan direction and to demonstrate the performances of the
different scanning methods previously described.
Finally, the reported results only concern topographical
characterization. With the same kind of setup, additional information can potentially be retrieved. We are currently
working on the integration of the SHU to a combined atomic
force-scanning near-field optical microscope.2 Further works
will also include adapting the SHU to ultimate applications
such as lithography using a scanning electron microscope
共SEM兲. The SEM is, in fact, very constraining since it requires very high speeds, temperature stabilization in addition
to operating in vacuum. For this, reducing the SHU volume
is currently under study.
ACKNOWLEDGMENTS
The authors are grateful toward E. Oseret from the Laboratoire de Parallélisme, Réseaux, Systèmes, et Modélisation
共PRiSM兲 for his useful discussions about handling high density images. This project is cofunded by the French Government 共ACI Nanosciences兲 and by the Ile-de-France region.
1
FIG. 5. 共Color online兲 Zoom of one of the waveguide structure. It is obtained directly from the first image without scanning once again. The upper
part presents a 3D view in order to show the waveguide profile, and the
lower part presents a cross section which allows proper measurements. The
figure illustrates one waveguide but statistics could be done over a set of
structures.
The International Technology Roadmap for Semiconductors, Metrology
report, 2005, http://www.itrs.net/Common/2005ITRS/Metrology2005.pdf
R. Bachelot, G. Lerondel, S. Blaize, S. Aubert, A. Bruyant, and P. Royer,
Microsc. Res. Tech. 64, 441 共2004兲.
3
S. Blaise, S. Aubert, A. Bruyant, R. Bachelot, G. Lerondel, P. Royer, J. E.
Broquin, and V. Minier, J. Microsc. 209, 588 共2003兲.
4
S. C. Minne, J. D. Adams, G. Yaralioglu, S. R. Manalis, A. Atalar, and C.
2
Downloaded 03 Jun 2008 to 193.51.41.209. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp
095107-7
Enlarged AFM scanning scope
F. Quate, Appl. Phys. Lett. 73, 1742 共1998兲.
H. Liu, B. Lu, Y. Ding, Y. Tang, and D. Li, J. Micromech. Microeng. 13,
295 共2003兲.
6
J. Fu, R. D. Young, and T. V. Vorburger, Rev. Sci. Instrum. 63, 2200
共1992兲.
7
H.-C. Yeh, W.-T. Ni, and S.-S. Pan, Control Eng. Pract. 13, 559 共2005兲.
8
M. Holmes, R. Hocken, and D. Trumper, Precis. Eng. 24, 191 共2000兲.
9
J. A. Kramar, Meas. Sci. Technol. 16, 2121 共2005兲.
10
G. Wilkening and L. Koenders, Nanoscale Calibration Standards and
Methods, 共Wiley-VCH, New York, 2005兲.
11
F. Meli and R. Thalmann, Meas. Sci. Technol. 9, 1087 共1998兲.
12
J. Haycocks and K. Jackson, Precis. Eng. 29, 168 共2005兲.
13
R. Leach, J. Haycocks, K. Jackson, A. Lewis, S. Oldfield, and A. Yaccot,
Nanotechnology 12, R1 共2001兲.
14
S. Ducourtieux, S. Duhem, F. Larsonnier, J. Salgado, G. P. Vailleau, L.
Lahousse, and J. David, in Proceedings of the 5th Euspen International
Conference, Montpellier, France, 共2005兲 Vol. 1, pp. 131–134 .
15
C.-W. Lee and S.-W. Kim, Precis. Eng. 21, 113 共1997兲.
16
J. Schneir, T. H. McWaid, J. Alexander, and B. P. Wilfley, J. Vac. Sci.
Technol. B 12, 3561 共1994兲.
17
M. Bienias, S. Gao, K. Hasche, R. Seemann, and K. Thiele, Appl. Phys.
A: Mater. Sci. Process. 66, 837 共1998兲.
18
G. B. Picotto and M. Pisani, Ultramicroscopy 86, 247 共2001兲.
5
Rev. Sci. Instrum. 78, 095107 共2007兲
19
S. Gonda, T. Doi, T. Kurosawa, Y. Tanimura, N. Hisata, T. Yamagishi, H.
Fujimoto, and H. Yukawa, Rev. Sci. Instrum. 70, 3362 共1999兲.
J. Haycocks and K. Jackson, in Proceedings of the 2nd Euspen
Conference, Torino, Italy, 2001, p. 392.
21
K. R. Koops and K. Dirscherl, in Proceedings of 3rd Euspen Conference,
Eindhoven, The Netherlands, 2002 , pp. 525–528.
22
W. Gao, Y. Arai, A. Shibuya, S. Kiyono, and C. H. Park, Precis. Eng. 30,
96 共2006兲.
23
S. Topçu, L. Chassagne, D. Haddad, Y. Alayli, and P. Juncar, Rev. Sci.
Instrum. 74, 4876 共2003兲.
24
S. Topçu, L. Chassagne, Y. Alayli, and P. Juncar, Opt. Commun. 247, 133
共2005兲.
25
L. Chassagne, S. Topcu, Y. Alayli, and P. Juncar, Meas. Sci. Technol. 16,
1771 共2005兲.
26
K. Dirscherl, Ph.D. thesis, Technical University of Denmark, Lyngby,
2000.
27
C.-L. Hwang, Y.-M. Chen, and C. Jan, IEEE Trans. Control Syst. Technol.
13, 56 共2005兲.
28
N. Bobroff, Precis. Eng. 15, 33 共1993兲.
29
F. C. Demarest, Meas. Sci. Technol. 9, 1024 共1998兲.
30
Pacific Nanotechnology Inc., Nano-R atomic force microscope specifications, http://www.pacificnanotech.com/nano-r-spm.html
31
K. P. Birch and M. J. Downs, Metrologia 30, 155 共1993兲.
20
Downloaded 03 Jun 2008 to 193.51.41.209. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp